Giving eyes to pelagic trawls: Acoustic and optical techniques measure behaviour, species, and sizes of fish in situ

Abstract

Trawling, towing a cone-shaped net behind a moving boat, is a widespread fishing method both in commercial fisheries and to collect fish for scientific investigations. It combines filtering effect with herding behaviour of fish in response to the vessel and components of the trawl to concentrate them in its path. For pelagic trawls, designed to be fished in the water column with little or no contact on the seabed, mesh openings in the forward and belly sections can be metres across making them relatively inefficient filters and reliant on herding to guide fish into the codend where meshes are small enough to prevent fish from escaping. Nevertheless, few studies have focused on the behaviour of fish, particularly large gadoids, during pelagic trawling and as a result trawl designs and fishing strategies are likely not optimized either for commercial harvest or research sampling. The first investigation described in this thesis revealed that shoaling Atlantic cod (Gadus morhua) captured in a commercial fishing setting using a pelagic trawl dove following vessel passage and swam towards the approaching trawl, with a significant proportion of fish escaping beneath the trawl. Once they were inside the trawl, the cod turned and swam slowly in the direction of trawling but were carried deeper into the trawl by its greater speed through water. They remained in the lower portion of the trawl, suggesting the top panel played little role in retaining fish and could be modified to reduce drag without reducing the catch. Despite the use of three acoustic sensors and multiple mechanical catch sensors mounted to the trawl, poor information was available during trawling on the species, sizes, and quantity of fish entering and already inside the trawl. In a commercial fishery, this would likely result in bycatch and discards. In response to this information gap, an in-trawl camera system, DeepVision, was developed to identify and measure all fish as they passed into the codend. Stereo photogrammetric techniques were developed to calculate lengths of fish from the images, and counts by species and individual fish lengths match well with standard physical sampling of the catch. The system has generated significant interest within the fisheries research community as a tool to provide enhanced information during research trawling and was tested during an annual ecosystem survey in the Norwegian Sea. Benefits shown during this trial included documentation of fine-scale patterns in spatial distribution by species and sizes, documentation of external parasites on Atlantic cod and saithe (Pollachius virens), and positive identification of the species and size composition of acoustically visible layers. The mis-identification of whiting (Merlangius merlangius) as pollack (Pollachius pollachius) during routine catch sampling was uncovered during review of the DeepVision data, demonstrating its value as a tool to quality check data even after the catch has been discarded. The technique opens the possibility to reduce sampling mortality while still registering large numbers of fish by trawling with an open codend or in conjunction with multisampler equipment to collect small, directed, biological samples for physical analysis.